Recherche avancée

Sur d’autres sites (6561)

• rtsp : Don’t warn about unparsed time ranges

  22 avril 2015, par Martin Storsjö

  rtsp : Don’t warn about unparsed time ranges
  This removes the error logging added in 4e54432164.
  This avoids warnings about "Invalid interval start specification ’now’"
  
  for live rtsp streams.
  We only try to parse some of the many valid values for time ranges
  
  in RTSP - the other ones are fully valid but not interesting for the
  
  use case in rtsp.c, so we shouldn’t warn about them.
  (Parsing the time ranges is needed to allow seeking, but e.g. setting
  
  the current realtime clock for the start time doesn’t make sense.
  
  av_parse_time has got a different mode for parsing absolute times
  
  as well, which can handle the special case "now", but that doesn’t
  
  make much sense for this particular use in rtsp.c.)
  Signed-off-by : Martin Storsjö <martin@martin.st>
  

  • [DH] libavformat/rtsp.c

• ffmpeg - Seek to absolute time stamp in MPEG DASH segment [closed]

  30 mai 2024, par iBent

  I need to extract short audio segments at specific time stamps from specific dash segments.

  


  I tried the following :

  


  ffmpeg -ss 00:10:00 -i segment_x.m4s -t 10 out.mp3


  


  -ss seeks relative to the segment's start time however, and not absolutely.
That absolute time data is there however as ffmpeg prints it during the conversation like this :

  


  Duration: 00:10:14.01, start: 595.018667, bitrate: 7 kb/s


  


  How can I make ffmpeg extract the audio from exactly 00:10:00 to 00:10:10 ?

  


• Can -axcorrelate be used for time delay analysis ?

  31 octobre 2020, par koral

  I'm trying to find a way to easily automatically synchronize two audio tracks. I've found the axcorrelate filter in ffmpeg's documentation, and I'm curious if it can be used for time delay analysis with cross-correlation.

  


  I've been playing around with correlating different audio files and opening them in Audacity to see some sort of a pattern that can be detected in the output file, but I wasn't able to see anything that stands out to me.

  


  My observations :

  


  

  • correlating two same tracks results in mostly empty output

  


  • if I delay one of the input files by one second, the correlation start (time in wav file when for the first time a sound is output) is moved by one second forward, but the delay cannot be calculated since the original start time is unknown

  


  • correlating some original file with that original file delayed by some amount results in gibberish data, that seems random to me

  


  


  Any help would be appreciated. I found this pdf that uses Praat to detect the cross correlation, but I'd like to keep this FFmpeg only to not ship multiple dependencies to my users.